The official title of the Amazon Science blog is “Alexa gets better at predicting customers’ goals,” but you should read this as, “The Turing test is not for machines to pass, but for humans to fail.”
I’m excited and terrified that the nice people at Alexa Voice Services are getting closer to a system improvement that would have the AI model infer the intent of a question you ask, then ask you a subsequent question.
The example Amazon offers: “Alexa… how long does it take to steep tea?” The system would answer your question, then follow up with, “Would you like to set a timer?”
I find myself getting annoyed at my already-too-verbose Alexa. (Although I have it set to speak with a female British accent, so it’s hard to get mad at “her.”) I’m taking this too lightly, which is a mistake. This capability is not only impressive, but exceptionally useful.
Over time, the model will understand what you are asking in a way that you will not. As it is trained by millions of people, this kind of model will speed research and help us get to the answers we seek, which may modify “If you ask the wrong question, you are guaranteed to get the wrong answer” to “If you ask the wrong question, Alexa will ask you the right one.”
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it.